1CLINFO(1) General Commands Manual CLINFO(1)
2
6 clinfo - show OpenCL platforms and devices
7
10 clinfo [options ...]
11
14 clinfo prints all available information about all OpenCL platforms
15 available on the system and the devices they expose.
16
19 clinfo accepts the following options:
20 -a, --all-props
22 try to retrieve all properties, even those not officially supported
23 (e.g. because they require specific extensions), but only show them
24 if the property could be retrieved successfully; see also the LIMITA‐
25 TIONS section below; note that even though this may reveal hidden
26 properties, there is no guarantee that the returned values are mean‐
27 ingful, nor that the corresponding feature is actually available at
28 all;
29 -A, --always-all-props
31 like -a, but also show errors;
32 --human
34 produce human-friendly output; this is the default (except as noted
35 below);
36 --raw
38 produce machine-friendly output; this is the default if clinfo is in‐
39 voked with a name that contains the string “raw”;
40 --json
42 outputs the raw data (cf. the --raw option) in JSON format; support
43 for this option is experimental, as the representation of some of the
44 values is not finalized;
45 --offline
47 shows also offline devices for platforms that expose this feature;
48 -l, --list
50 list platforms and devices by name, with no (other) properties;
51 -dplatform_index:device_index
53 --device platform_index:device_index
54 only show properties for the specified device in the specified plat‐
55 form;
56 --prop property-name
58 only show properties whose symbolic name matches (contains as a sub‐
59 string) the given property-name; the name is normalized as upper-case
60 and with minus sign (-) replaced by underscore signs (_); when this
61 flag is specified, raw mode is forced;
62 -h, -?
64 show usage;
65 --version, -v
67 show program version.
68
71 OpenCL 1.1, OpenCL 1.2, OpenCL 2.0, OpenCL 2.1, OpenCL 2.2, OpenCL 3.0.
72
75 Supported OpenCL extensions:
76 cl_khr_device_uuid
78 for the UUID, LUID and node mask of the device;
79 cl_khr_extended_versioning
81 for the extended platform, device, extension and IL versioned proper‐
82 ties backported from OpenCL 3.0 to previous OpenCL versions;
83 cl_khr_fp16, cl_khr_fp64, cl_amd_fp64, cl_APPLE_fp64_basic_ops
85 for information about support for half-precision and double-precision
86 floating-point data types;
87 cl_khr_image2d_from_buffer
89 for information about the base address and pitch alignment require‐
90 ments of buffers to be used as base for 2D images;
91 cl_khr_il_program
93 for information about the supported IL (Intermediate Language) repre‐
94 sentations;
95 cl_khr_spir
97 for information about the supported SPIR (Standard Portable Interme‐
98 diate Representation) versions;
99 cl_khr_icd
101 for the suffix of vendor extensions functions;
102 cl_khr_subgroup_named_barrier
104 for the maximum number of named sub-group barriers;
105 cl_khr_terminate_context
107 for the terminate capabilities for the device;
108 cl_ext_device_fission
110 for device fission support in OpenCL 1.1 devices;
111 cl_ext_atomic_counters_32
113 cl_ext_atomic_counters_64
114 for the atomic counter extension;
115 cl_ext_cxx_for_opencl
117 for the version of the C++ for OpenCL language supported by the de‐
118 vice compiler;
119 cl_amd_device_attribute_query
121 for AMD-specific device attributes;
122 cl_amd_object_metadata
124 to show the maximum number of keys supported by the platform;
125 cl_amd_offline_devices
127 to show offline devices exposed by the platform, if requested (see
128 --offline option);
129 cl_amd_copy_buffer_p2p
131 to show the number and IDs of available P2P devices;
132 cl_amd_svm
134 cl_arm_shared_virtual_memory
135 for Shared Virtual Memory (SVM) capabilities in OpenCL 1.2 devices;
136 cl_arm_core_id
138 to show the (potentially sparse) list of the core IDs that the device
139 may return;
140 cl_arm_job_slot_selection
142 to show the (potentially sparse) list of available job slots for com‐
143 mand submission;
144 cl_nv_device_attribute_query
146 for NVIDIA-specific device attributes;
147 cl_intel_exec_by_local_thread
149 for the Intel extension allowing CPU devices to run kernels as part
150 of the current host thread;
151 cl_intel_advanced_motion_estimation
153 for the version of the Intel Motion Estimation accelerator version;
154 cl_intel_device_side_avc_motion_estimation
156 for the version and supported features of Intel's device-side AVC Mo‐
157 tion;
158 cl_intel_planar_yuv
160 for the maximum dimensions of planar YUV images;
161 cl_intel_simultaneous_sharing
163 for simultaneous CL/GL/DirectX context sharing (only partial sup‐
164 port);
165 cl_intel_required_subgroup_size
167 to enumerate allowed sub-group sizes;
168 cl_altera_device_temperature
170 for the Altera extension to query the core temperature of the device;
171 cl_qcom_ext_host_ptr
173 for the QUALCOMM extension to query page size and required padding in
174 external memory allocation.
175
178 Some information is duplicated when available from multiple sources.
179 Examples:
180 • supported device partition types and domains as obtained using the
182 cl_ext_device_fission extension typically match the ones obtained us‐
183 ing the core OpenCL 1.2 device partition feature;
184 • the preferred work-group size multiple matches the NVIDIA warp size
186 (on NVIDIA devices) or the AMD wavefront width (on AMD devices).
187 Some floating-point configuration flags may only be meaningful for spe‐
190 cific precisions and/or specific OpenCL versions. For example,
191 CL_FP_CORRECTLY_ROUNDED_DIVIDE_SQRT is only relevant for single preci‐
192 sion in OpenCL 1.2 devices.
193 The implementation-defined behavior for NULL platform or context prop‐
196 erties is tested for the following API calls:
197 clGetPlatformInfo()
199 by trying to show the platform name;
200 clGetDeviceIDs()
202 by trying to enumerate devices; the corresponding platform (if any)
203 is then detected by querying the device platform of the first device;
204 clCreateteContext()
206 by trying to create a context from a device from the previous list
207 (if any), and a context from a device from a different platform;
208 clCreateteContextFromType()
210 by trying to create contexts for each device type (except DEFAULT).
211
214 Support for OpenCL 2.x properties is not fully tested.
215 Support for cl_khr_subgroup_named_barrier is experimental due to miss‐
218 ing definitions in the official OpenCL headers.
219 Raw (machine-parsable) output is considered experimental, the output
222 format might still undergo changes.
223 The properties of the ICD loader will also be queried if the clGetICD‐
226 LoaderInfoOCLICD extension function is found.
227 Support for the properties exposed by cl_amd_copy_buffer_p2p is experi‐
230 mental.
231 Support for some (documented and undocumented) properties exposed by
234 cl_amd_device_attribute_query is experimental (see also LIMITATIONS).
235 Support for the interop lists exposed by cl_intel_simultaneous_sharing
238 is experimental.
239 The highest OpenCL version supported by the ICD loader is detected with
242 some trivial heuristics (symbols found); a notice is output if this is
243 lower than the highest platform OpenCL version, or if the detected ver‐
244 sion doesn't match the one declared by the ICD loader itself.
245
248 OpenCL provides no explicit mean to detect the supported version of any
249 extension exposed by a device, which makes it impossible to determine a
250 priori if it will be possible to successfully query a device about a
251 specific property. Additionally, the actual size and meaning of some
252 properties are not officially declared anywhere.
253 Most notably, this affects extensions such as cl_amd_device_attri‐
256 bute_query, cl_nv_device_attribute_query and cl_arm_core_id. Heuris‐
257 tics based on standard version support are partially used in the code
258 to determine which version may be supported.
259 Properties which are known to be affected by these limitations include:
262 CL_DEVICE_GLOBAL_FREE_MEMORY_AMD
265 documented in v3 of the cl_amd_device_attribute_query extension spec‐
266 ification as being the global free memory in KBytes, without any ex‐
267 planation given on why there are two values, although in the source
268 code of the ROCm stack the second value is documented as being the
269 largest free block;
270 CL_DEVICE_AVAILABLE_ASYNC_QUEUES_AMD
272 documented in v3 of the cl_amd_device_attribute_query extension spec‐
273 ification, but not reported by drivers supporting other v3 proper‐
274 ties. This has now been enabled for drivers assumed to support v4 of
275 the same extension;
276 CL_DEVICE_TERMINATE_CAPABILITY_KHR
278 exposed by the cl_khr_terminate_context has changed value between
279 OpenCL 1.x and 2.x, and it's allegedly a bitfield, whose values are
280 however not defined anywhere.
281
284 General
285 Please report any issues on the project tracker on GitHub
286 ⟨http://github.com/Oblomov/clinfo⟩.
287 LLVM CommandLine errors
290 If multiple OpenCL platforms using shared LLVM libraries are present in
291 the system, clinfo (and other OpenCL application) may crash with errors
292 to the tune of
293 : CommandLine Error: Option '(some option name)' registered more than once!
295 LLVM ERROR: inconsistency in registered CommandLine options
296 or similar. This is not an issue in clinfo, or in any OpenCL platform
298 or application, but it is due to the way LLVM handles its own command-
299 line options parsing. The issue has been reported upstream as issue
300 #30587 ⟨https://bugs.llvm.org/show_bug.cgi?id=30587⟩. See the next
301 point for possible workarounds and assistance in identifying the con‐
302 flicting platforms.
303 Segmentation faults
306 Faulty OpenCL platforms may cause segmentation faults in clinfo during
307 the information gathering phase, sometimes even before any output is
308 shown. There is very little clinfo can do to avoid this. If you see
309 this happening, try disabling all platforms and then re-enabling them
310 one by one until you experience the crash again. Chances are the last
311 platform you enabled is defective in some way (either by being incom‐
312 patible with other platforms or by missing necessary components and not
313 handling their absence gracefully).
314 To selectively enable/disable platforms, one way is to move or rename
317 the *.icd files present in /etc/OpenCL/vendors/ and then restoring them
318 one by one. When using the free-software ocl-icd OpenCL library, a sim‐
319 ilar effect can be achieved by setting the OPENCL_VENDOR_PATH or
320 OCL_ICD_VENDORS environment variables, as documented in libOpenCL(7).
321 Other implementations of libOpenCL are known to support OPENCL_VEN‐
322 DOR_PATH too.
323 Example
326 find /etc/OpenCL/vendors/ -name '*.icd' | while read OPENCL_VEN‐
327 DOR_PATH ; do clinfo -l > /dev/null ; echo "$? ${OPENCL_VENDOR_PATH}"
328 ; done
329 This one liner will run clinfo -l for each platform individually (hid‐
332 ing the normal output), and report the .icd path prefixed by 0 for suc‐
333 cessful runs, and a non-zero value for faulty platforms.
334clinfo 3.0.21.02.21 2021-02-21 CLINFO(1)